I may have finally found a clean solution to the issues[1][2][3] with
the automatic discovery of wireless network interfaces.
[1]: https://github.com/NixOS/nixpkgs/issues/101963
[2]: https://github.com/NixOS/nixpkgs/issues/23196
[3]: https://github.com/NixOS/nixpkgs/pull/125917#issuecomment-856000426
Currently the start script fails right away if no interface is available
by the time it's running, possibly leaving the system without network.
This happens when running a little early in the boot. A solution is to
instead wait for at least one interface to appear before scanning the
/sys/class/net/ directory. This is done here by listening for the right
udev events (from the net/wlan subsystem) using the `udevadm monitor`
command and grep to match its output.
This methods guarantees the availability of at least one interface to
wpa_supplicant, but won't add additional interfaces once it has started.
However, if the current interface is lost, say unplugged, the service is
automatically stopped and will be restarted as soon as a one (not
necessarily the same) is detected. It would be possible make this fully
dynamic by running another service that continously listen for udev
events and manages the main wpa_supplicant daemon, but this is probably
overkill.
I tested the following cases:
- one interface, starting at boot, w/o predictable naming scheme
- two interfaces, starting at boot (intel wireless and a usb adapter),
w/o predictable naming scheme
- one interface after the system booted, w/o predictable naming scheme
- two interfaces after the system booted, w/o predictable naming scheme
- unplugging and plugging back the current interface
The pppd daemon starting with version 2.4.9 uses rtnetlink to configure
the ipv6 peer address on the ppp interface. It therefore requires
allowing AF_NETLINK sockets.
The kernel before version 5.7 required CAP_SYS_ADMIN to conduct BPF
operations. After that a separate capability CAP_BPF was created, which
should be sufficient in this scenario and will further tighten the
sandbox around our pppd service.
Tested on my personal DSL line.
- The order of NSS (host) modules has been brought in line with upstream
recommendations:
- The `myhostname` module is placed before the `resolve` (optional) and `dns`
entries, but after `file` (to allow overriding via `/etc/hosts` /
`networking.extraHosts`, and prevent ISPs with catchall-DNS resolvers from
hijacking `.localhost` domains)
- The `mymachines` module, which provides hostname resolution for local
containers (registered with `systemd-machined`) is placed to the front, to
make sure its mappings are preferred over other resolvers.
- If systemd-networkd is enabled, the `resolve` module is placed before
`files` and `myhostname`, as it provides the same logic internally, with
caching.
- The `mdns(_minimal)` module has been updated to the new priorities.
If you use your own NSS host modules, make sure to update your priorities
according to these rules:
- NSS modules which should be queried before `resolved` DNS resolution should
use mkBefore.
- NSS modules which should be queried after `resolved`, `files` and
`myhostname`, but before `dns` should use the default priority
- NSS modules which should come after `dns` should use mkAfter.
This allows users of the bind module to specify an alternate BIND
package. For example, by overriding the source attribute to use a
different version of BIND.
Since the default value for `services.bind.package` is `pkgs.bind`,
this change is completely backwards compatible with the current
module.
